What happens when an AI model “thinks”?

Read [Alex Duffy’s “We Trained an AI on a Board Game. It Became a Better Customer Support Agent”](https://every.to/playtesting/we-trained-an-ai-on-a-board-game-it-became-a-better-customer-support-agent-299b5938-09dd-4881-803f-aea21f0d461f) as one example of how training shapes a model’s reasoning. Then help me understand how reasoning models work and why they helped make agents like Codex possible. Make me an accessible explainer with one concrete example. Explain how training, time spent working through a problem, and access to tools each contribute. Help me understand what “reasoning” means here and where the analogy to human thinking breaks down. Use reliable sources, distinguish established explanations from uncertainty, and end with two questions we could explore next. Assume I’m curious but not an AI researcher.